[Foundation-l] Freedom, standards, and file formats

[Gregory Maxwell]

On Mon, Sep 29, 2008 at 12:16 AM, Tim Starling <tstarling at wikimedia.org> wrote:
[snip]
> I'm suggesting Flash because I believe that education, not promotion of
> free software, should be our goal. We should only support free software as
> far as it supports that goal. As a community, we believe that free
> software supports our goal a great deal, and that we should use it
> everywhere where it is practical.

Where did free software come into this discussion.  It's orthogonal,
almost entirely so.  The formats we use have BSD licensed reference
implementations and are supported in a great many pieces of
proprietary software.

I very much wish a particular piece of proprietary software, Microsoft
Windows, integrated Ogg/Theora support because then I very much think
the discussion of flash video would be over. (Microsoft ships Xiph
codecs in many other products... so it's less unreasonable than you
might guess, though even Adobe ships Ogg/Speex in Flash, but I'm told
that Theora has no chance in flash today).

I use proprietary software to convert proprietary formats into ones
suitable for Wikipedia.  You don't hear me howling that the office
uses many Mac (though I did like it when Danny had Ubuntu(?) on his
desktop).

I think we're in agreement on the role of free software in Wikimedia
But formats are not software.  At the time I first setup the Java
video player for Wikimedia, Java itself was still fairly proprietary
(though I did get the player working in GCJ first).   If Flash played
free media formats and you were to propose to use flash *only* for
that purpose, I would not pose an objection.  (though for other
purpose, flash has other problems).

Our mission is not merely education, but educational content under
free content licenses. This much is clear, unambiguously stated, and
not currently up for debate. It's a key differentiation between
Wikimedia and hundreds of thousands of other educators.

Those licenses grant rights -- like the ability to create derivatives
or verbatim distribution at no cost and without asking for permission
-- which are not realistically possible in a world where using a
no-cost format means that only computer geeks can view the result, or
where you have to republish via a site like Wikimedia who will pay
some of the licensing fees for you.

> Various organisations have distributed Wikipedia text in non-free formats,
> and I hope we continue to encourage that.

I'd hope that we'd encourage them to use free formats where at all
possible, while still doing whatever good thing they are doing.  But
they are not us, and what they do is their concern, not ours.

Depending on the exact nature of their non-free formats and their
usage that action may be in violation of the licensing: Producing DRM
locked enhanced versions is expressly forbidden by both the FDL and by
the CC-By-SA family of licenses.   I would hope that Wikimedia is wise
enough to not encourage any such illegal and arguably unethical
activity.

[snip]
> You argue that we would have to exclude more people in the future, as a
> consequence of Theora losing ground compared to where it is now. I think
> you are overestimating this effect.

It doesn't have to lose ground: Our perspective on costs will change.
After a couple years of flash video, do you really think that going
back to where we are now would be an equally easy decision than we
have now in staying the course?

As Kat's post states:  Formats enjoy tremendous network effects.   For
things with strong network effect we expect logistic or exponential
like growth curves:  Investing $10 today is not the same as investing
$10 ten years from now.

Our video support sucks in a dozen ways unrelated to not offering
non-free video.  The added cost of telling people to download another
codec/player is not huge compared to the other issues. Once that
changes the argument will be different.  Internet users of today are
very accustomed to installing software: If the iPhone is at all
predictive, we may simply not have the same opportunities in the
future.

I may well be overestimating the effect our use will have in the
future,  but I know my past expectations were underestimates.

Someone has to pay a cost to get the free formats adopted:  What this
argument is really fundamentally about is "Shall we take a gamble and
try to externalize that cost?"

Of all the large websites we're the least likely to tolerate non-free
formats, since we actually see non-freeness itself as a cost. Many
other large sites actually gain an advantage from the non-freeness.
What does that say about the probability of others picking up that
cost if we do not?  I do not think it is a good gamble at all.

I don't even think that the cost is all that great:  In a frictionless
market we would expect the codec prices to singular large
organizations (like Wikimedia), or at least co-operating collections
(too bad that almost all other big media sites are either MPEG patent
holders or in brutal competition with each other) to be roughly
comparable to the cost of spurring adoption.  I'm not aware of any
studies of this, but its easy to see the drop in competing audio codec
licensing fees that happened when Ogg/Vorbis went 1.0.  It's probably
even less for us because the pricing was probably designed for groups
with a decreased incompatibility tolerance.   ... Though I guess by
letting OS vendors pay the decoder fees, and Free Software
users/vendors by via legal liability, downstream users by their need
to pay royalties, we probably can externalize the majority of the cost
associated with a decision to use non-free formats.

... Enough waxing economic.

>> Do you actually think the difference in quality per bitrate between
>> the current Theora encoder and H.264 has any relevance to us?
>
> Not directly, but technical parity might help to convince vendors to
> support it.

Perhaps. But do you not think that "Wikipedia requires it?" would not
be a much more  effective argument (at least for a web-browsing
device).

I've talked to hardware makers about format support in the past, and
it's all about market size. I really don't think anything else matters
at all: Even fees.  If H.264 cost $25/device (I think the original
licensing needed DVD players were that high), and H.264 is required by
the market, they'll simply pass the cost on, since all the major
players will be compelled to pay the same fees.

Codec licensing normally has per-company annual caps, so the largest
companies with the most ability to cause adoption actually see a
competitive advantage vs smaller competition caused by the non-free
formats that they have to pay for.

> Granted. But on principle, I don't want the lack of support for commercial
> client software to be written into our bylaws.

Fantastic! because no one has proposed that.  Seriously.

Please don't confuse free formats with free software. Freely licensed
software needs free formats. Freely licensed content needs free
formats.  But Freely licensed formats have advantages to even to
people who care nothing about either of those two things.

The free codecs we use today have BSD licensed reference
implementations (a direct recommendation by RMS, in fact, for the
express purpose of encouraging adoption by proprietary software).
Free media formats are used in many pieces proprietary software, and
shipped by the largest proprietary software vendor in the world
(though sadly not with their web browser).  Flash includes Speex,
mostly likely due to Speex's through domination of a particular market
segment.